Bug#721298: boinc-client: Idle Detection Not Working Computer Always In Use

Tue Dec 15 23:34:22 UTC 2015

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA512

An update:

Re-incorporating the XSS idle-detection code works as expected. I
scrapped the DPMS code and only kept the XSS code. Inclusion of

#include <X11/extensions/scrnsaver.h>

in "client/hostinfo_unix.cpp" ultimately pulls in a conflicting
definition of "Always" between X-related code and boinc's
"lib/prefs.h", fixed by renaming boinc's declaration. After this, the
boinc-client and boincmgr compile and idle detection works as expected
(keyboard and mouse inputs are appropriately detected, and computation
is suspended).

Working on cleaning up the changes, hopefully getting upstream to
reinstate the XSS code, and determining how best to get the fix into
the Debian packages. From what I can tell, an additional dependency on
"libxss1" would be necessary.

Cheers,
Preston

On 12/14/2015 07:28 AM, Preston Maness wrote:
> Whelp, it looks like the X-based idle detection --with the 
> XScreenSaver API?-- was removed in this commit:
> 
> ==================
> 
> bfae1032e5c1ac73f2d8d92f222293d8383a6cee
> 
> https://github.com/BOINC/boinc/commit/bfae1032e5c1ac73f2d8d92f222293d8
38
>
> 
3a6cee
> 
> ==================
> 
> The commit message indicates that /dev/input/* devices are the 
> intended way to determine idle time on Linux now. However, it
> seems that X-based inputs don't update the /dev/input/* devices. Or
> at least, anything the X server is doing isn't updating the atime
> on those files, which is what is being utilized to determine if
> the system is idle or not. Working in a terminal updates the atime
> on the relevant files --correctly triggering BOINC to suspend
> computation-- but just typing this email, or wiggling the mouse
> around the desktop area, doesn't.
> 
> Granted, this code was present earlier, so I don't know if simply 
> re-introducing it will solve anything. But that's my first plan of 
> attack. I've also looked into other varying methods of idle
> detection. The XSS API is the most common, but there is a more
> convoluted (asynchronous?) Xsync option that might also be possible
> too. I don't know too much about X, but I'll be giving a few
> different possibilities a go in the coming days.
> 
> Cheers, Preston
> 
> On 12/11/2015 09:44 AM, Preston Maness wrote:
>> Howdy howdy,
> 
>> I think I deserve a slow clap for such a slow
>> response/addresssing of a bug. Indeed my response is so slow that
>> I've since changed email address and GPG keys.
> 
>> Needless to say, when running BOINC 7.6.12 from testing, some 
>> progress seems to have been made on idle detection. If I open a 
>> terminal and begin working in it, *sometimes*, the idle detection
>>  kicks in. I can't tell whether it's working as intended or not.
>>  Regardless, keeping boinc running non-stop doesn't pose any
>> issues thanks to the scheduling. I can happily use my machine for
>> just about anything and not even notice that BOINC is running.
> 
>> I'm presently unemployed, so really, I have nothing but time and
>> no real excuses not to give this bug another go and see if I can
>> just determine once and for all if things are fixed/how to fix
>> them :P
> 
>> It looks like the BOINC project has moved from their own git 
>> infrastructure onto github:
> 
>> https://boinc.berkeley.edu/trac/wiki/SourceCodeGit
> 
>> https://github.com/BOINC/boinc
> 
>> Looking at the git log for hostinfo_unix.cpp (a la "git log 
>> client/hostinfo_unix.cpp" from the project root), it looks like 
>> there was a commit that included changes from Gianfranco:
> 
>> ==================
> 
>> commit 0811886cbb1d282376e2dfd017ab39e34cd7d8ea Author: David 
>> Anderson <davea at ssl.berkeley.edu> Date:   Mon Feb 3 12:46:43 2014
>>  -0800
> 
>> client: fix for idle detection on Linux (from Gianfranco)
> 
> 
>> ==================
> 
>> and some later refactoring of the unix HOSTINFO code:
> 
>> ==================
> 
>> commit cda6bd334bf036349f471a0c4eb7c25de8098262 Author: David 
>> Anderson <davea at ssl.berkeley.edu> Date:   Tue Nov 25 14:46:58
>> 2014 -0800
> 
>> client: get static host info items only on startup
> 
>> On Android, the way we were getting os_version (Linux version + 
>> Android version) didn't work because get_host_info() was getting
>>  called before every scheduler RPC, and it overwrote the Android
>>  version part.
> 
>> Solution: divide host info into dynamic (disk usage, network
>> info) and static (everything else). Computer the static part only
>> at startup.
> 
>> Also factor the Unix HOST_INFO code into multiple functions.
> 
>> ==================
> 
>> As it stands, BOINC's Event Log does show periodic "Suspending 
>> computation - computer is in use" messages. Show I'm operating on
>>  the assumption that something is working.
> 
>> Cheers, Preston Maness
> 
> 
> 

-----BEGIN PGP SIGNATURE-----
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=AW/A
-----END PGP SIGNATURE-----